It works as well as any other block set up that requires no block updates (like candlesticks, flowerpots, etc.) and is only a risk if a builder griefs which we have a good record of. I'm not particularly concerned about block updates.
Yes, those aren't good either, and I'd very much like to move them onto custom blocks. The difference is these blocks became used in this manner after they were already present in the game, whereas now we have a choice to either add these crops as custom blocks or purposefully add them in an unstable and hacky way.
Builders accidentally place blocks when they're exploring all the time. I remember the water in the KL sewers being broken a multitude of times before we finally figured out the other water layer ID doesn't break. It's painstaking to fix and usually takes a while to detect, as sometimes the builder doesn't even notice it happen.
We've also known for a long time that WE'd half-banners and half-beds can cause memory leaks (I believe illblew pointed that out a while ago). I have no idea if this is also the case with these types of unstable blocks, but it seems like it's a possibility. Again, I'd rather err on the safe side when we have a choice to do so with little inconvenience to us.
I don't think is particularly necessary, new crops don't outdate builds as much as for example a new block set, especially since up until now we've used "representational" stand ins of the crops I mentioned that can pass for other crops or variations of them.
I disagree that it wouldn't outdate things. What's the point of adding these new crops if they're just substitutes for the existing stand-ins in terms of quality? There is the issue of ambiguous block uses for future RP functionality, certainly, but (a) that idea also seems to outdated old builds, and almost resigns every one to be redone eventually unless there's targeted efforts to update them, and (b) we're already between a rock and a hard place in that regard due to how many of our blocks are used this way. It's hard to tell what'll happen when we get to that point, but based on the UE4 experiments it seems like there's a chance we won't even be relying on MC functionality for that effort, in which case the point is moot.